Re: [PATCH] riscv: put va_kernel_xip_pa_offset into CONFIG_XIP_KERNEL

[Conor Dooley]

On Wed, Dec 20, 2023 at 10:14:59PM +0100, Alexandre Ghiti wrote:
> Hi Yunhui,
> 
> On 20/12/2023 11:34, Yunhui Cui wrote:
> > opitmize the kernel_mapping_pa_to_va() and kernel_mapping_va_to_pa().
> > 
> > Signed-off-by: Yunhui Cui <cuiyunhui@xxxxxxxxxxxxx>
> > ---
> >   arch/riscv/include/asm/page.h | 33 ++++++++++++++++++++-------------
> >   1 file changed, 20 insertions(+), 13 deletions(-)
> > 
> > diff --git a/arch/riscv/include/asm/page.h b/arch/riscv/include/asm/page.h
> > index 5488ecc337b6..0d2b479d02cd 100644
> > --- a/arch/riscv/include/asm/page.h
> > +++ b/arch/riscv/include/asm/page.h
> > @@ -113,8 +113,8 @@ struct kernel_mapping {
> >   	unsigned long va_pa_offset;
> >   	/* Offset between kernel mapping virtual address and kernel load address */
> >   	unsigned long va_kernel_pa_offset;
> > -	unsigned long va_kernel_xip_pa_offset;
> >   #ifdef CONFIG_XIP_KERNEL
> > +	unsigned long va_kernel_xip_pa_offset;
> >   	uintptr_t xiprom;
> >   	uintptr_t xiprom_sz;
> >   #endif
> > @@ -134,12 +134,25 @@ extern phys_addr_t phys_ram_base;
> >   #else
> >   void *linear_mapping_pa_to_va(unsigned long x);
> >   #endif
> > -#define kernel_mapping_pa_to_va(y)	({					\
> > -	unsigned long _y = (unsigned long)(y);					\
> > -	(IS_ENABLED(CONFIG_XIP_KERNEL) && _y < phys_ram_base) ?			\
> > -		(void *)(_y + kernel_map.va_kernel_xip_pa_offset) :		\
> > -		(void *)(_y + kernel_map.va_kernel_pa_offset + XIP_OFFSET);	\
> > -	})
> > +
> > +#ifdef CONFIG_XIP_KERNEL
> > +#define kernel_mapping_pa_to_va(y)							\
> > +	(((unsigned long)(y) < phys_ram_base) ?						\
> > +		(void *)((unsigned long)(y) + kernel_map.va_kernel_xip_pa_offset) :	\
> > +		(void *)((unsigned long)(y) + kernel_map.va_kernel_pa_offset + XIP_OFFSET))
> > +
> > +#define kernel_mapping_va_to_pa(y)						\
> > +	(((unsigned long)(y) < kernel_map.virt_addr + XIP_OFFSET) ?		\
> > +		((unsigned long)(y) - kernel_map.va_kernel_xip_pa_offset) :	\
> > +		((unsigned long)(y) - kernel_map.va_kernel_pa_offset - XIP_OFFSET))
> > +#else
> > +#define kernel_mapping_pa_to_va(y)						\
> > +	((void *)((unsigned long)(y) + kernel_map.va_kernel_pa_offset + XIP_OFFSET))
> > +
> > +#define kernel_mapping_va_to_pa(y)						\
> > +	((unsigned long)(y) - kernel_map.va_kernel_pa_offset - XIP_OFFSET)
> > +#endif
> > +
> >   #define __pa_to_va_nodebug(x)		linear_mapping_pa_to_va(x)
> >   #ifndef CONFIG_DEBUG_VIRTUAL
> > @@ -147,12 +160,6 @@ void *linear_mapping_pa_to_va(unsigned long x);
> >   #else
> >   phys_addr_t linear_mapping_va_to_pa(unsigned long x);
> >   #endif
> > -#define kernel_mapping_va_to_pa(y) ({						\
> > -	unsigned long _y = (unsigned long)(y);					\
> > -	(IS_ENABLED(CONFIG_XIP_KERNEL) && _y < kernel_map.virt_addr + XIP_OFFSET) ? \
> > -		(_y - kernel_map.va_kernel_xip_pa_offset) :			\
> > -		(_y - kernel_map.va_kernel_pa_offset - XIP_OFFSET);		\
> > -	})
> >   #define __va_to_pa_nodebug(x)	({						\
> >   	unsigned long _x = x;							\
> 
> 
> Not sure using #ifdef optimizes anything since the compiler should do the
> same with the IS_ENABLED(CONFIG_XIP_KERNEL) and it does not really improve
> the readability of this file which is already overloaded with #ifdef, so I
> don't think this change is needed.

I would say that we explicitly do not want to move things that are
guarded by IS_ENABLED() to ifdeffery. In fact, we should move things in
the other direction if possible, especially for stuff like XIP_KERNEL
that nobody ever build tests.

Attachment: signature.asc
Description: PGP signature